答题卡判卷实战
在实现答题卡判卷中,有部分操作与文档OCR识别重合,分别是预处理、轮廓提取、透视变换。
难点为透视变换。
新学习内容部分为答案判定部分。
获取轮廓(轮廓检测)后得到答题卡轮廓中的四个点。
def order_points(pts):
# 一共4个坐标点
rect = np.zeros((4, 2), dtype = "float32")
# 按顺序找到对应坐标0123分别是 左上,右上,右下,左下
# 计算左上,右下
s = pts.sum(axis = 1)
rect[0] = pts[np.argmin(s)]
rect[2] = pts[np.argmax(s)]
# 计算右上和左下
diff = np.diff(pts, axis = 1)
rect[1] = pts[np.argmin(diff)]
rect[3] = pts[np.argmax(diff)]
return rect
关于边缘检测:
(1)获取图像
(2)ROI裁剪(目前没用到。一般resize),滤波前将图像转为灰度图。如果颜色特征对你的研究没用,只需要形状亮度等特征,就需要转化为 灰度图 。 灰度图需要处理的数据比彩色图小很多,处理更快,也没有颜色干扰。
(3)图像滤波
(4)提取边缘,threshold(阈值处理),还有不同算子
(5)边缘处理
(6)显示结果
透视变换:
首先需要获取提取出的轮廓的四个点,计算两条长宽(w、h)的值,分别留下最大的那个使用,得到最终的四个坐标点,利用M矩阵进行透视变换。
def four_point_transform(image, pts):
# 获取输入坐标点(得到的四个点)
rect = order_points(pts)
(tl, tr, br, bl) = rect
# 计算输入的w和h值
widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
maxWidth = max(int(widthA), int(widthB))
heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
maxHeight = max(int(heightA), int(heightB))
# 变换后对应坐标位置(用数组存储对应的四个坐标点)
dst = np.array([
[0, 0],
[maxWidth - 1, 0],
[maxWidth - 1, maxHeight - 1],
[0, maxHeight - 1]], dtype = "float32")
# 计算变换矩阵
M = cv2.getPerspectiveTransform(rect, dst)
warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
# 返回变换后结果
return warped
排序操作:
def sort_contours(cnts, method="left-to-right"):
reverse = False
i = 0
if method == "right-to-left" or method == "bottom-to-top":
reverse = True
if method == "top-to-bottom" or method == "bottom-to-top":
i = 1
boundingBoxes = [cv2.boundingRect(c) for c in cnts]
(cnts, boundingBoxes) = zip(*sorted(zip(cnts, boundingBoxes),
key=lambda b: b[1][i], reverse=reverse))
return cnts, boundingBoxes
整体流程:
答题卡识别判卷:遍历轮廓–透视变换(形态学处理)–二值处理–遍历轮廓–找到每一个圆圈的轮廓–遍历排序圆圈–mask填充判断结果(通过计算非零点(这是灰度为0的意思)的数量(白色255的地方?)判断是否选择了这个图案)(mask掩码与操作)–对比正确答案–计算得分
完整代码
#导入工具包
import numpy as np
import imutils
import cv2
# 正确答案
ANSWER_KEY = {0: 1, 1: 4, 2: 0, 3: 3, 4: 1}
def order_points(pts):
# 一共4个坐标点
rect = np.zeros((4, 2), dtype = "float32")
# 按顺序找到对应坐标0123分别是 左上,右上,右下,左下
# 计算左上,右下
s = pts.sum(axis = 1)
rect[0] = pts[np.argmin(s)]
rect[2] = pts[np.argmax(s)]
# 计算右上和左下
diff = np.diff(pts, axis = 1)
rect[1] = pts[np.argmin(diff)]
rect[3] = pts[np.argmax(diff)]
return rect
def four_point_transform(image, pts):
# 获取输入坐标点
rect = order_points(pts)
(tl, tr, br, bl) = rect
# 计算输入的w和h值
widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
maxWidth = max(int(widthA), int(widthB))
heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
maxHeight = max(int(heightA), int(heightB))
# 变换后对应坐标位置
dst = np.array([
[0, 0],
[maxWidth - 1, 0],
[maxWidth - 1, maxHeight - 1],
[0, maxHeight - 1]], dtype = "float32")
# 计算变换矩阵
M = cv2.getPerspectiveTransform(rect, dst)
warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
# 返回变换后结果
return warped
def sort_contours(cnts, method="left-to-right"):
reverse = False
i = 0
if method == "right-to-left" or method == "bottom-to-top":
reverse = True
if method == "top-to-bottom" or method == "bottom-to-top":
i = 1
boundingBoxes = [cv2.boundingRect(c) for c in cnts]
(cnts, boundingBoxes) = zip(*sorted(zip(cnts, boundingBoxes),
key=lambda b: b[1][i], reverse=reverse))
return cnts, boundingBoxes
def cv_show(name,img):
cv2.imshow(name, img)
cv2.waitKey(0)
cv2.destroyAllWindows()
# 预处理
image = cv2.imread('F:/BaiduNetdiskDownload/test_05.png')
contours_img = image.copy()
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blurred = cv2.GaussianBlur(gray, (5, 5), 0)
cv_show('blurred',blurred)
edged = cv2.Canny(blurred, 75, 200)
cv_show('edged',edged)
# 轮廓检测
cnts = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)[0]
contours_img = image.copy()
print(len(cnts))
cv2.drawContours(contours_img,cnts,-1,(0,0,255),3)
cv_show('contours_img',contours_img)
docCnt = None
# 确保检测到了
if len(cnts) > 0:
# 根据轮廓大小进行排序
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)
# 遍历每一个轮廓
for c in cnts:
# 近似
peri = cv2.arcLength(c, True)
approx = cv2.approxPolyDP(c, 0.02 * peri, True)
# 准备做透视变换
if len(approx) == 4:
docCnt = approx
break
# 执行透视变换
warped = four_point_transform(gray, docCnt.reshape(4, 2))
cv_show('warped',warped)
# Otsu's 阈值处理
thresh = cv2.threshold(warped, 0, 255,
cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1]
cv_show('thresh',thresh)
thresh_Contours = thresh.copy()
# 找到每一个圆圈轮廓
cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)[0]
cv2.drawContours(thresh_Contours,cnts,-1,(0,0,255),3)
cv_show('thresh_Contours',thresh_Contours)
questionCnts = []
# 遍历
for c in cnts:
# 通过做外接矩形,计算外接矩形长宽的大小和比例,如果比例为1或接近于1则代表是个圆形
(x, y, w, h) = cv2.boundingRect(c)
ar = w / float(h)
# 根据实际情况指定标准(有误差)
if w >= 20 and h >= 20 and ar >= 0.9 and ar <= 1.1:
questionCnts.append(c)
# 按照从上到下进行排序(圆圈)
questionCnts = sort_contours(questionCnts,
method="top-to-bottom")[0]
correct = 0
# 同一道题的选项,x轴值相同y轴值不同
# 每排有5个选项(△)
for (q, i) in enumerate(np.arange(0, len(questionCnts), 5)):
# 排序
cnts = sort_contours(questionCnts[i:i + 5])[0]
bubbled = None
# 遍历每一个结果
for (j, c) in enumerate(cnts):
# 使用mask(掩码)(与操作)来判断结果
mask = np.zeros(thresh.shape, dtype="uint8")
cv2.drawContours(mask, [c], -1, 255, -1) #-1表示填充
cv_show('mask',mask)
# 通过计算非零点数量来算是否选择这个答案
# 此处的mask是该选项的实际值(计算像素点的差异?)
# 执行与操作(255+255/255+0?)
mask = cv2.bitwise_and(thresh, thresh, mask=mask)
total = cv2.countNonZero(mask)
# 通过阈值判断,找到当前题目不为0的所选答案
if bubbled is None or total > bubbled[0]:
bubbled = (total, j)
# 对比正确答案
color = (0, 0, 255)
k = ANSWER_KEY[q]
# 判断正确
if k == bubbled[1]:
# 绿色
color = (0, 255, 0)
correct += 1
# 绘图
cv2.drawContours(warped, [cnts[k]], -1, color, 3)
score = (correct / 5.0) * 100
print("[INFO] score: {:.2f}%".format(score))
cv2.putText(warped, "{:.2f}%".format(score), (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 0, 255), 2)
cv2.imshow("Original", image)
cv2.imshow("Exam", warped)
cv2.waitKey(0)